§ MENU
Approach
§ PRODUCTS · PLATINE Platine Vibe OS § PLATFORMS WE LEAD ON Dynamics 365 Business Central Project Online · Migration
§ THE PRACTICE Engagements Industries § THINKING & WRITING Field Notes
About Contact → Français → Book a call
INDUSTRY CASE STUDY · Q4 2025 · 5-MIN READ

The 991-app problem, what the connectivity data shows.

MuleSoft's annual benchmark and Forrester's integration TEI agree: the average enterprise is more disconnected than its leadership realizes, and the cost compounds quietly.

TL;DR
  • The average enterprise runs 991 distinct applications. Only 28% of them are integrated with each other.
  • Knowledge workers spend roughly 30% of their time on data movement and reconciliation that integrated systems eliminate.
  • AI initiatives in disconnected estates fail at roughly 3× the rate of integrated peers, integration is now the prerequisite for higher-order digital programs.

The Salesforce / MuleSoft 2024 Connectivity Benchmark Report, surveying more than 1,000 IT leaders across enterprise organizations, published the number that quietly defines the operating reality of most large companies: 991 distinct applications running on average, with only 28% of them connected to each other.

The disconnection is not just a technical inconvenience. The downstream costs compound through every operational layer.

The cost of disconnection, in research

IDC and McKinsey have separately quantified what the gap costs in productivity. Knowledge workers across surveyed enterprises spend roughly 30% of their working time searching for, reconciling, or moving data between disconnected systems. For a 500-person organization, that is the equivalent of 150 full-time employees doing nothing but middleware work, without any of the architectural rigor of an actual integration platform.

Forrester's Total Economic Impact studies on enterprise integration platforms consistently show that closing the gap returns 200–300% over three years, not from headcount reduction, but from decisions made faster, errors caught earlier, and dashboards that finally reflect operating reality.

The integration → AI dependency

The most striking finding of the 2024 cycle came from Gartner: through 2026, AI initiatives in organizations without strong data integration foundations will fail at roughly 3× the rate of integrated peers. The reason is mechanical, large language models are only as useful as the data they can access, and data scattered across hundreds of unconnected applications is data the model cannot ground its answers in.

This reframes integration as something different from infrastructure plumbing. Integration is now the prerequisite for every higher-order digital initiative, AI, automation, real-time analytics, customer-360 work. The order of operations matters: integrate first, AI second.

What survives the disconnection

The minority of organizations whose integration estates work well share two traits across the studies. First, they treat integration as a sustained operating discipline rather than a project, with documented architectures, monitoring, and clear ownership. Second, they prefer fewer, well-instrumented integrations over many fragile point-to-point connections. The 28% that work are the ones that were built to be operated, not just shipped.

SOURCES  ·  Salesforce / MuleSoft, "2024 Connectivity Benchmark Report"  ·  Gartner, 2024 integration & AI research  ·  Forrester Total Economic Impact studies on enterprise iPaaS  ·  IDC / McKinsey Global Institute knowledge-worker productivity research

§, RELATED

More from the field.

§ FN · INTERROGATIVE

Want to discuss a similar engagement?

Schedule
→ Book a 30-minute call
All notes
→ Browse all field notes
Phone
514-546-0711